==System Environment== -------------------------- Regression: not sure Non-working platforms: BYT ==kernel== -------------------------- drm-intel-nightly/bd21cf795cf5dc278f1451d0f7a597cb1d13c2ba ==Bug detailed description== Run 2 cycles, it happens once. output: root@x-byt06:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./drv_hangman --run-subtest error-state-capture-bsd IGT-Version: 1.8-gab5f7ea (x86_64) (Linux: 3.18.0-rc2_drm-intel-nightly_bd21cf_20141030+ x86_64) Subtest error-state-capture-bsd: SUCCESS (12.504s) root@x-byt06:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# dmesg -r|egrep "<[1-4]>"|grep drm root@x-byt06:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./drv_hangman --run-subtest error-state-capture-bsd IGT-Version: 1.8-gab5f7ea (x86_64) (Linux: 3.18.0-rc2_drm-intel-nightly_bd21cf_20141030+ x86_64) Subtest error-state-capture-bsd: SUCCESS (12.057s) root@x-byt06:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# dmesg -r|egrep "<[1-4]>"|grep drm <3>[19432.618274] [drm:__vlv_force_wake_get [i915]] *ERROR* Timed out: waiting for Render to ack. dmesg: [19408.408488] drv_hangman: executing [19408.408671] drv_hangman: starting subtest error-state-capture-bsd [19414.574594] [drm] stuck on bsd ring [19414.582570] [drm] GPU HANG: ecode 1:0xfffffffe, in drv_hangman [9123], reason: Ring hung, action: reset [19414.582574] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace. [19414.582576] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel [19414.582578] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue. [19414.582581] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it. [19414.582583] [drm] GPU crash dump saved to /sys/class/drm/card0/error [19414.585604] [drm] Simulated gpu hang, resetting stop_rings [19414.585610] drm/i915: Resetting chip after gpu hang [19426.892311] drv_hangman: executing [19426.892526] drv_hangman: starting subtest error-state-capture-bsd [19432.610176] [drm] stuck on bsd ring [19432.618152] [drm] GPU HANG: ecode 1:0xfffffffe, in drv_hangman [9131], reason: Ring hung, action: reset [19432.618274] [drm:__vlv_force_wake_get [i915]] *ERROR* Timed out: waiting for Render to ack. [19432.620510] [drm] Simulated gpu hang, resetting stop_rings [19432.620516] drm/i915: Resetting chip after gpu hang ==Reproduce steps== ---------------------------- 1. ./drv_hangman --run-subtest error-state-capture-bsd
gem_reset_stats also has this issue. @test: Intel_gpu_tools/igt_gem_reset_stats_reset-count-blt returncode: 0 info: @@@Returncode: 0 test case start at: Thu Oct 30 12:32:10 2014 test case end at: Thu Oct 30 12:32:20 2014 Errors: Dmesg: <6>[ 108.951070] [drm] stuck on blitter ring <6>[ 108.958779] [drm] GPU HANG: ecode 2:0xe77ffff3, in gem_reset_stats [5973], reason: Ring hung, action: reset <6>[ 108.958787] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace. <6>[ 108.958789] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel <6>[ 108.958792] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue. <6>[ 108.958794] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it. <6>[ 108.958797] [drm] GPU crash dump saved to /sys/class/drm/card0/error <3>[ 108.958906] [drm:__vlv_force_wake_get [i915]] *ERROR* Timed out: waiting for Render to ack. <6>[ 108.960558] [drm] Simulated gpu hang, resetting stop_rings <5>[ 108.960564] drm/i915: Resetting chip after gpu hang
Some kms_flip case also has this issue. @test: Intel_gpu_tools/igt_kms_flip_vblank-vs-hang-interruptible returncode: 0 info: @@@Returncode: 0 test case start at: Sun Nov 2 17:03:29 2014 test case end at: Sun Nov 2 17:04:15 2014 Errors: Dmesg: <6>[ 82.897180] [drm] stuck on render ring <6>[ 82.905149] [drm] GPU HANG: ecode 0:0x97f4fffe, in kms_flip [4654], reason: Ring hung, action: reset <6>[ 82.905153] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace. <6>[ 82.905155] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel <6>[ 82.905157] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue. <6>[ 82.905160] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it. <6>[ 82.905162] [drm] GPU crash dump saved to /sys/class/drm/card0/error <6>[ 82.906904] [drm] Simulated gpu hang, resetting stop_rings <5>[ 82.906910] drm/i915: Resetting chip after gpu hang <6>[ 91.914805] [drm] stuck on render ring <6>[ 91.922612] [drm] GPU HANG: ecode 0:0x97f4fffe, in kms_flip [4654], reason: Ring hung, action: reset <3>[ 91.922791] [drm:__vlv_force_wake_get [i915]] *ERROR* Timed out: waiting for Render to ack. <6>[ 91.924479] [drm] Simulated gpu hang, resetting stop_rings <5>[ 91.924486] drm/i915: Resetting chip after gpu hang <6>[ 100.936360] [drm] stuck on render ring <6>[ 100.944273] [drm] GPU HANG: ecode 0:0x97f4fffe, in kms_flip [4654], reason: Ring hung, action: reset <6>[ 100.946063] [drm] Simulated gpu hang, resetting stop_rings <5>[ 100.946069] drm/i915: Resetting chip after gpu hang <6>[ 108.947934] [drm] stuck on render ring <6>[ 108.955718] [drm] GPU HANG: ecode 0:0x97f4fffe, in kms_flip [4654], reason: Ring hung, action: reset <6>[ 108.957675] [drm] Simulated gpu hang, resetting stop_rings <5>[ 108.957680] drm/i915: Resetting chip after gpu hang
It also impacts BSW. [root@x-bsw01 tests]# ./drv_hangman --run-subtest error-state-basic IGT-Version: 1.8-gab5f7ea (x86_64) (Linux: 3.18.0-rc2_drm-intel-nightly_203b34_20141103+ x86_64) Subtest error-state-basic: SUCCESS (15.659s) [root@x-bsw01 tests]# dmesg -r|egrep "<[1-4]>"|grep drm <3>[ 235.778310] [drm:__vlv_force_wake_get [i915]] *ERROR* Timed out: waiting for Render to ack.
Please 'cat /sys/module/i915/parameters/enable_ppgtt' and paste result. Does this happen if you use i915_enable.ppgtt=1?
# cat /sys/module/i915/parameters/enable_ppgtt 0 Add i915_enable.ppgtt=1, it still happens. igt_gem_workarounds_reset also has this error. Dmesg: <6>[ 8678.817582] [drm] stuck on render ring <6>[ 8678.824135] [drm] GPU HANG: ecode 0:0x97f4fffe, in gem_workarounds [20809], reason: Ring hung, action: reset <6>[ 8678.824143] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace. <6>[ 8678.824146] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel <6>[ 8678.824148] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue. <6>[ 8678.824151] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it. <6>[ 8678.824153] [drm] GPU crash dump saved to /sys/class/drm/card0/error <3>[ 8678.824242] [drm:__vlv_force_wake_get [i915]] *ERROR* Timed out: waiting for Render to ack. <6>[ 8678.826343] [drm] Simulated gpu hang, resetting stop_rings <5>[ 8678.826347] drm/i915: Resetting chip after gpu hang
(In reply to lu hua from comment #5) > # cat /sys/module/i915/parameters/enable_ppgtt > 0 > > Add i915_enable.ppgtt=1, it still happens. > i915.enable_ppgtt=1
Created attachment 108880 [details] [review] Wait old forcewake ack to clear on vlv
*** Bug 85358 has been marked as a duplicate of this bug. ***
(In reply to Mika Kuoppala from comment #7) > Created attachment 108880 [details] [review] [review] > Wait old forcewake ack to clear on vlv Test this patch 10 cycles, it doesn't happen.
Apparently a regresion due to > > commit 5cb13c07dae73380d8b3ddc792740487b8742938 > > Author: Deepak S <deepak.s@linux.intel.com> > > Date: Thu Sep 18 18:51:50 2014 +0530 > > > > drm/i915/vlv: Remove check for Old Ack during forcewake mark it as such.
Test it on latest -nightly kernel 10 cycles, it works well. root@x-bytm02:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./drv_hangman --run-subtest error-state-capture-bsd IGT-Version: 1.8-gc049c39 (x86_64) (Linux: 3.18.0-rc3_drm-intel-nightly_b921a5_20141110+ x86_64) Subtest error-state-capture-bsd: SUCCESS (13.560s) root@x-bytm02:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# dmesg -r|egrep "<[1-4]>"|grep drm
Created attachment 109259 [details] dmesg(gem_reset_stats) Test all gem_reset_stats subcases on BYT, it still has this error. root@x-byt06:/GFX/Test/Infrastructure/infrastructure# dmesg -r|egrep "<[1-4]>"|grep drm <3>[ 162.062426] [drm:__vlv_force_wake_get [i915]] *ERROR* Timed out: waiting for Render to ack. <3>[ 237.209028] [drm:__vlv_force_wake_get [i915]] *ERROR* Timed out: waiting for Render to ack. <3>[ 249.232659] [drm:__vlv_force_wake_get [i915]] *ERROR* Timed out: waiting for Render to ack. <3>[ 505.732108] [drm:__vlv_force_wake_get [i915]] *ERROR* Timed out: waiting for Render to ack. Intel_gpu_tools/igt_gem_reset_stats_ban-blt DMESG_WARN reboot Intel_gpu_tools/igt_gem_reset_stats_ban-ctx-blt NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_ban-ctx-bsd NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_ban-ctx-render PASS Intel_gpu_tools/igt_gem_reset_stats_ban-ctx-vebox NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_ban-render PASS Intel_gpu_tools/igt_gem_reset_stats_ban-vebox NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_close-pending-blt DMESG_WARN reboot Intel_gpu_tools/igt_gem_reset_stats_close-pending-bsd DMESG_WARN reboot Intel_gpu_tools/igt_gem_reset_stats_close-pending-ctx-blt NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_close-pending-ctx-bsd NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_close-pending-ctx-render PASS Intel_gpu_tools/igt_gem_reset_stats_close-pending-ctx-vebox NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_close-pending-fork-blt PASS Intel_gpu_tools/igt_gem_reset_stats_close-pending-fork-bsd PASS Intel_gpu_tools/igt_gem_reset_stats_close-pending-fork-render PASS Intel_gpu_tools/igt_gem_reset_stats_close-pending-fork-reverse-blt PASS Intel_gpu_tools/igt_gem_reset_stats_close-pending-fork-reverse-bsd PASS Intel_gpu_tools/igt_gem_reset_stats_close-pending-fork-reverse-render PASS Intel_gpu_tools/igt_gem_reset_stats_close-pending-fork-reverse-vebox NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_close-pending-fork-vebox NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_close-pending-render PASS Intel_gpu_tools/igt_gem_reset_stats_close-pending-vebox NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_params PASS Intel_gpu_tools/igt_gem_reset_stats_params-ctx-blt NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_params-ctx-bsd NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_params-ctx-render PASS Intel_gpu_tools/igt_gem_reset_stats_params-ctx-vebox NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_reset-count-blt PASS Intel_gpu_tools/igt_gem_reset_stats_reset-count-bsd PASS Intel_gpu_tools/igt_gem_reset_stats_reset-count-ctx-blt NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_reset-count-ctx-bsd NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_reset-count-ctx-render PASS Intel_gpu_tools/igt_gem_reset_stats_reset-count-ctx-vebox NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_reset-count-render PASS Intel_gpu_tools/igt_gem_reset_stats_reset-count-vebox NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_reset-stats-blt DMESG_WARN reboot Intel_gpu_tools/igt_gem_reset_stats_reset-stats-bsd PASS Intel_gpu_tools/igt_gem_reset_stats_reset-stats-ctx-blt NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_reset-stats-ctx-bsd NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_reset-stats-ctx-render PASS Intel_gpu_tools/igt_gem_reset_stats_reset-stats-ctx-vebox NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_reset-stats-render PASS Intel_gpu_tools/igt_gem_reset_stats_reset-stats-vebox NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_unrelated-ctx-blt NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_unrelated-ctx-bsd NSPT reboot Intel_gpu_tools/igt_gem_reset_stats_unrelated-ctx-render PASS Intel_gpu_tools/igt_gem_reset_stats_unrelated-ctx-vebox NSPT reboot
Run drv_hangman, the error still exists. root@x-byt06:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./drv_hangman IGT-Version: 1.8-gc049c39 (x86_64) (Linux: 3.18.0-rc3_drm-intel-nightly_f56717_20141105+ x86_64) Subtest error-state-debugfs-entry: SUCCESS (0.000s) Subtest error-state-sysfs-entry: SUCCESS (0.000s) Subtest ring-stop-sysfs-entry: SUCCESS (0.001s) Subtest error-state-basic: SUCCESS (6.057s) Subtest error-state-capture-render: SUCCESS (7.766s) Subtest error-state-capture-bsd: SUCCESS (12.350s) Subtest error-state-capture-blt: SUCCESS (19.028s) Test requirement not met in function gem_require_ring, file ioctl_wrappers.c:881: Test requirement: gem_has_vebox(fd) Subtest error-state-capture-vebox: SKIP (0.000s) root@x-byt06:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# dmesg -r|egrep "<[1-4]>"|grep drm <3>[ 6621.654382] [drm:__vlv_force_wake_get [i915]] *ERROR* Timed out: waiting for Render to ack.
Offending commit has been reverted: commit 9500986159d9017f9277f9e42a1f9b13b5e0e666 Author: Mika Kuoppala <mika.kuoppala@linux.intel.com> Date: Wed Nov 5 17:30:52 2014 +0200 Revert "drm/i915/vlv: Remove check for Old Ack during forcewake"
Verified on latest -nightly(3cb89f9eef2888c848248bf45d6dd0d67c594586) root@x-byt06:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./drv_hangman --run-subtest error-state-capture-bsd IGT-Version: 1.8-gaa63fc7 (x86_64) (Linux: 3.18.0-rc5_kcloud_3cb89f_20141119+ x86_64) Subtest error-state-capture-bsd: SUCCESS (12.330s) root@x-byt06:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# dmesg -r|egrep "<[1-4]>"|grep drm root@x-byt06:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests#
Closing old verified+fixed.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.